default

What Changes When AI Enters Production in Travel

May 14, 2026
3 Min read
What Changes When AI Enters Production in Travel
Share it!

Building on the PhocusWire article “What it takes to make AI work at scale in travel,” this Q&A brings Gaëlle Bristiel, Vice President of Engineering at Amadeus, back into the conversation to examine what changes in practice when AI operates across production‑grade travel systems.



AI has been part of travel technology for decades. What has changed is the context in which it now operates. As AI moves into live, end‑to‑end travel systems, it becomes subject to the same operational demands as core booking and servicing processes.


In travel, these environments allow little margin for error. Systems are deeply interconnected, decisions flow across multiple entities, and reliability is essential to keep operations running smoothly. Once AI is embedded into these workflows, the focus shifts from model performance to execution, control and trust in live operational conditions.

What changes in day‑to‑day operations once AI runs in live travel systems?

Once AI runs in live travel systems, success is no longer measured by outputs alone.


In practice, AI becomes part of travel systems that have a near zero tolerance for error, where reliable and consistent behavior must align with core booking and servicing flows that support day‑to‑day operations.

Why does orchestration become critical once AI is deployed in production environments?

In travel, AI decisions often span multiple systems, entities and workflows.


Orchestration is required to sequence actions, align decisions with authoritative systems of record, and maintain traceability across end‑to‑end processes. This ensures that AI operates as part of a controlled system rather than in isolation.

Where do specialized AI agents typically face limitations in travel environments?

Specialized AI agents perform well within a clearly defined task or domain. Limitations emerge at the boundaries between tasks, where context or responsibility must be transferred.


When decisions need to span multiple tasks, a lack of clear handoffs can lead to fragmented outcomes across workflows.


At scale, addressing these limitations also depends on how AI assistants access and act on dynamic content across the travel ecosystem, within trusted and resource‑efficient frameworks.

How do travel companies test and build trust in non‑deterministic AI systems?

Beyond coordination and system design, production‑grade AI raises a more fundamental challenge: how trust is established and maintained in systems that do not behave deterministically.


In practice, AI systems in travel are validated and trusted through a combination of testing, monitoring and observability:


  • Testing focuses on ranges of acceptable responses, rather than exact matches.
  • Observability in production is essential to understand how AI behaves over time.
  • Continuous monitoring helps detect inconsistencies early.

Operating AI reliably at scale also requires clear responsibility, governance and accountability at an organizational level.


At Amadeus, we take a responsible approach to innovation and AI. This includes sustained investment in cybersecurity, transparency, compliance and resilience to ensure trust is built into everything we develop.

What role does Amadeus play in enabling AI to operate reliably at scale in travel?

Amadeus enables AI to operate reliably at scale by acting as a neutral execution layer connected to trusted travel systems of record. This anchors AI interactions in authoritative booking and servicing data and allows them to be coordinated across multiple stakeholders within production‑grade platforms, rather than layered on top of existing systems.


Amadeus also works closely with major technology partners that influence how travelers access travel services, ensuring these tools connect to the industry’s operational reality.


These capabilities are further reinforced through investments such as the acquisition of SkyLink, which strengthens orchestration and enables conversational booking and servicing within governed, large‑scale travel environments.

TO TOP

TO TOP